How hybrid infrastructure is making quantum commercially actionable today – and why the organizations building that infrastructure now will be the ones ready when the hardware arrives.
Insider Brief
- Quantum readiness is an infrastructure problem as much as a hardware one: organizations are in a “prepare for quantum” era, and integrating isolated quantum processors with enterprise IT through secure, low-latency interconnection is the step most organizations are overlooking.
- Paris-based Qubit Pharmaceuticals shows this approach in practice. Its four-layer ATLAS platform combines polarizable molecular dynamics, a synthetic quantum chemistry data engine built with Sorbonne University, the FeNNix-Bio1 foundation model, and the Hyperion-1 emulator, which simulates up to 40 qubits on classical hardware.
- Quantum advantage has not yet been demonstrated in drug discovery, but the milestones that matter now (algorithm validation, workflow integration and ecosystem maturity) don’t require fault-tolerant hardware. Organizations that delay preparation risk falling further behind as a result.
The Era Between Eras
The prevailing assumption about quantum computing is that fault-tolerant hardware will arrive at some point, organizations will integrate it when it does, and until then there is nothing practical to do.
That framing is both inaccurate and increasingly costly to believe. We are not in a “wait for quantum” era. We are in a “prepare for quantum” era, and preparation is an active, infrastructure-intensive undertaking. Organizations that treat it as passive will find themselves starting from scratch when quantum utility arrives. Those who build now will be able to absorb it.
The distinction matters most in the life sciences, where the cost of falling behind can be measured in years of delayed discovery for diseases that are not adequately served by today’s therapeutics. This is not a problem exclusive to drug discovery; it also affects finance, materials discovery, logistics optimization, and any domain where quantum is expected to eventually matter.
There is a meaningful difference between quantum science, the physics that governs how molecules, electrons, and systems behave, and enterprise-ready quantum computing, the applied, workload-integrated systems that can run inside real IT environments. Conflating the two leads organizations to either overestimate what is possible today or underestimate how much preparation is required. Both errors are costly.
So, the question of the hour is: what does preparation look like?
For starters, preparation is somewhat akin to multiplatform access. This means building the infrastructure necessary to access GPUs, QPUs, and whatever compute modality a workload may require. It is not about betting everything on a single path to quantum.
This case study, grounded in a concrete collaboration, clarifies and offers a window into the action when preparation meets practice.
“Our mission at Qubit Pharmaceuticals is to find molecules that can become groundbreaking treatments for diseases still considered incurable.”
— Qubit Pharmaceuticals
The Infrastructure Problem
Most quantum computing coverage focuses on qubits in terms of how many a given system has, how stable they are, or what error rates look like. These are real and important questions. But they are not the only bottleneck between where quantum is today and where enterprises need it to be.
The harder, less discussed problem is integration. Quantum hardware is geographically distributed and operationally isolated. Most systems live in specialized research environments, accessed via narrow APIs, disconnected from the classical compute, data pipelines, and security frameworks that enterprise IT depends on. Even organizations genuinely interested in quantum experimentation face a daunting challenge: how to connect a quantum processor to existing infrastructure in a way that is secure, low-latency, scalable, and doesn’t require rebuilding existing IT architecture from scratch.
This is the problem that neutral interconnection infrastructure is designed to solve, and it’s why the conversation about quantum readiness should begin with infrastructure, not solely with hardware specs.
Equinix’s position on this is direct: enterprise-ready quantum computing is less about the chip and more about the fabric connecting it. Hybrid quantum-classical architectures require quantum processors for intensive calculations and classical systems for data management to function as an integrated whole, not as separate tools accessed sequentially. That integration depends on secure, low-latency connectivity between quantum resources and enterprise IT infrastructure, available at the scale and geographic distribution that global organizations require.
| THE BOTTLENECK TODAY | THE INTERCONNECTION MODEL | THE OUTCOME |
| Quantum hardware is isolated from enterprise IT, geographically distributed, operationally fragmented, and difficult to integrate securely at scale. | A neutral infrastructure layer — low-latency, multi-cloud, ecosystem-connected — that lets classical, HPC, AI, and quantum workloads converge without rebuilding IT architecture. | Organizations can experiment with quantum capabilities today, build operational familiarity, and scale as hardware matures without a restart when the hardware arrives. |
Equinix Fabric and Network Edge are designed to provide software-defined connectivity enabling enterprises to integrate quantum capabilities with existing infrastructure, test proofs of concept before committing to full production, and scale usage as needs and hardware maturity evolve. The model treats quantum access the way cloud access has been treated for a decade – as a service reachable through the same interconnection layer as everything else, rather than as a bespoke on-premises installation.
“Enterprise-ready quantum computing isn’t about the chip. It’s about the fabric that connects it to everything else the business runs. ”
— Petrina Steele, Global Lead Emerging Technologies, Equinix
This infrastructure model delivers value before quantum hardware itself is production-ready. The same interconnection fabric that will eventually link quantum processors to enterprise workflows is today accelerating HPC, AI, and cloud-based simulation workloads. Building and operating on that fabric now is how you develop the operational fluency quantum will eventually require.
Qubit Pharmaceuticals: Hybrid Infrastructure in Action
Qubit Pharmaceuticals was founded in Paris in 2020 on the premise that quantum physics, properly applied, could make molecular simulation accurate enough to change how drugs are discovered. The company’s platform, ATLAS, is described as a “Quantum, AI and Physics-Inspired Platform for High-Precision Molecular Modeling” and is designed to make what the company calls “quantum accuracy” operationally available in a real discovery environment.
The interactions that determine whether a putative small molecule binds to its target, and with what strength and selectivity, are governed by quantum mechanical effects. Classical simulations rely on approximations of these effects, which can limit predictive accuracy. In drug discovery, those limits can surface later as candidates that look promising in simulation but fail experimentally, adding cost and time to each discovery cycle. Full quantum mechanical simulation would eliminate those approximations but is computationally prohibitive on classical hardware at biologically relevant scales.
Qubit’s response was not to wait for fault-tolerant quantum hardware. Qubit is building a platform architecture to systematically close the accuracy gap, using each available compute modality for what it does best today, while building readiness for quantum deployment as hardware matures.
The Architecture For Readiness
ATLAS is structured in four interlocking layers. The first is high-fidelity physics simulation: polarizable molecular dynamics and free-energy methods that are more computationally demanding than standard approaches but substantially more accurate, particularly in the water-rich biological environments where drug targets reside. The platform, which today runs on a dedicated 192-GPU cluster, already delivers throughput of 1,000 nanoseconds of molecular dynamics per day using polarizable force fields, a figure that matters because it determines how many candidates can be evaluated at meaningful accuracy within a real discovery timeline.
The second layer is a synthetic data engine that generates the training data that simulation alone can’t produce affordably. Working with Sorbonne University, Qubit built a large database of first-principles quantum chemistry calculations computed on major HPC systems, including GENCI, EuroHPC, and Argonne National Laboratory. That database feeds directly into the third layer.
The third layer is FeNNix-Bio1, a quantum AI foundation model announced by Qubit in May 2025. Trained entirely on the aforementioned synthetic quantum chemistry data, it achieves quantum-level accuracy in binding free energy calculations at a fraction of the compute cost. This is particularly useful for protein-ligand and hydration calculations, where simpler models fall short.
The fourth layer is Qubit’s quantum and quantum-hybrid algorithm program. Hyperion-1, a GPU-based emulator that can simulate up to 40 qubits on classical hardware, lets the team test and refine quantum algorithms before running them on real devices, and then validate them on available hardware as it scales up.
Built For What’s Coming
The platform architecture required deliberate infrastructure decisions at every level. For production compute, Qubit built on AWS, leveraging elastic scaling and the ability to run multiple discovery projects simultaneously without fixed hardware constraints. For GPU-accelerated simulation and hybrid programming, the platform draws on Nvidia’s systems and the CUDA-Q model for hybrid quantum-classical development, allowing the same codebase to target GPU clusters today and quantum-classical hybrid systems as hardware matures. For the performance-critical physics simulation work described above, Qubit also runs a dedicated 192-GPU cluster hosted at Equinix. This cluster orchestrates roughly 1.6 million GPU hours per year on it at a cost and utilization rate that a comparable hyperscaler footprint cannot match.
Splitting compute this way is possible thanks to Equinix’s cloud on-ramp: software-defined connectivity that lets Qubit move workloads between its private cluster and AWS rather than choosing one or the other. That flexibility supports data sovereignty and lowers storage costs for workloads that don’t need to run in the public cloud, while preserving the elasticity to burst into AWS when a project needs it. Where an organization lands on that spectrum, from fully cloud to hybrid, depends on how much infrastructure expertise it has in-house and how much it is already spending in the cloud. Qubit’s answer was to keep its constant, high-utilization compute on dedicated hardware it controls and to treat the cloud as an on-demand extension rather than the default.
For quantum access and ecosystem proximity, Qubit partnered with GENCI and HQI in France to have access to multiple modalities ranging from CPUs/GPUs to neutral atom, photonic, and eventually, superconducting quantum computers. The company also joined PINQ² within Quebec’s DistriQ quantum ecosystem, gaining access to an on-site IBM Quantum System One alongside HPC resources, and a physical presence within a cluster of quantum hardware providers, research institutions, and infrastructure partners. The logic mirrors what Equinix describes as the ecosystem proximity model, where proximity to quantum hardware providers and research partners maximizes the value of early experimentation and ensures infrastructure agility where it’s needed. Qubit is applying the same logic to quantum hardware itself: alongside its on-site IBM Quantum Platform through IBM Quantum Startups Program, a research collaboration with Singapore’s Centre for Quantum Technologies gives it experimental access to Quantinuum’s H2 through the Quantinuum Startup Partner Program and Helios quantum computers. Each is another platform in a portfolio that now spans GPUs, on-premises and cloud HPC, and multiple quantum hardware providers, so no single modality becomes a bottleneck as the technology matures.
“The Equinix Fabric infrastructure aims to interconnect Quantum Processing Units (QPUs) with classical High-Performance Computing (HPC) architectures (CPU/GPU) via a private network designed to minimize latency. This approach evolves quantum computing from an isolated cloud model (QCaaS) into an integrated and agnostic hybrid ecosystem. Architecturally, this coupling is considered fundamental for supporting hybrid molecular simulations applied to drug discovery, where the rapid, iterative exchange of information between classical and quantum resources is critical.”
— Jean-Philip Piquemal, CSO, Qubit Pharmaceuticals
What Hybrid Infrastructure Delivers Today
Quantum advantage, in the strict sense of a quantum processor outperforming any classical alternative on a commercially relevant task, has not yet been demonstrated in drug discovery. That benchmark remains ahead. What Qubit’s hybrid infrastructure delivers today is operational readiness, workflow integration, and capability building, all which compound over time.
| 83% | Faster progression from target to hit, driven by higher simulation throughput and fewer trial-and-error cycles on the Equinix-hosted GPU cluster. |
| 75% | Cost reduction versus industry-standard infrastructure, from running sustained, high-utilization GPU workloads on dedicated hardware rather than public cloud instances. |
| 40 qubits | Exact emulation scale achieved by Hyperion-1, enabling full algorithm validation on classical HPC before hardware deployment. This reduces wasted quantum machine time and accelerates development cycles. |
| 6+ | Active drug discovery programs across oncology, inflammation, antiviral, and RNA targets running on ATLAS in production, including partnered programs with Institut Curie and the University of Bordeaux. |
What Changes If This Scales
The Qubit case is instructive not because it is unique, but because it is replicable. The strategic logic in building accurate workflows, validating algorithms via emulation, building infrastructure to a quantum-ready specification, and connecting to quantum hardware as it matures is transferable to any domain where quantum is expected to eventually matter.
| FOR LIFE SCIENCES | FOR ENTERPRISE QUANTUM ADOPTION | FOR QUANTUM COMMERCIALIZATION |
| Multi-platform access, spanning GPU, AI, and QPU resources rather than any single modality, can accelerate earlier-stage discovery and open new target classes, such as complex RNA and protein targets where classical simulation struggles most. | Infrastructure-first strategy — building interconnection, HPC, and cloud integration before quantum hardware is production-ready — is the mechanism by which quantum readiness becomes operational rather than theoretical. | The ecosystem proximity model reduces barriers to experimentation, accelerates the path from proof-of-concept to production, and provides a template that other industries can follow as quantum as-a-service matures. |
Equinix has been clear about where this leads. Quantum computing will be realized through integration with existing IT infrastructure as a complement to classical infrastructure. Organizations that build toward that integration now will be better positioned to take advantage of both modalities as the technology matures. The complexity of that integration exceeds the technical capabilities of most organizations acting alone. Neutral interconnection infrastructure removes that complexity, making quantum accessible through the same fabric as cloud, HPC, and AI workloads.
What Comes Next
Several milestones define progress toward genuinely enterprise-ready quantum. One is algorithm validation at scale, demonstrated through emulation and early hardware tests on tasks that recur in production environments. Another is workflow integration, where quantum routines can be invoked as part of a larger classical pipeline without bespoke engineering each time. The final piece is ecosystem maturity, with quantum hardware, HPC, cloud, and research partners located on the same interconnection fabric, so that a new quantum capability can be accessed and tested without a procurement cycle.
None of these milestones require fault-tolerant quantum hardware to be reached first. They are already happening in organizations that chose to treat preparation as active rather than passive. Those who have not yet started are accumulating a compounding deficit, because the operational familiarity, algorithm libraries, and infrastructure integrations being built today will be prerequisites for quantum deployment tomorrow.
In the end, the organizations most likely to benefit from quantum computing are the ones that started building workflows, data infrastructure, algorithm libraries, and ecosystem connections before the hardware was ready to run them. As it stands today, for any organization operating in a quantum-relevant domain, it is not about whether to prepare, but how far behind they can afford to start.